Hi, I am a doctoral researcher at ETH Zürich and MPI-IS in Tübingen, advised by Celestine Mendler-Dünner (MPI) and Andreas Krause (ETH).

My research focuses on test-time training and alignment for large language models, with the goal of enhancing their adaptability and reliability in real-world settings. More broadly, I am interested in the theoretical foundations of machine learning and the development of robust, adaptive systems.

Projects

• Understanding Gradient Flow Dynamics for Matrix Factorization Problems

  Harvard & ETH Zürich, February 2024 – September 2024

  High-dimensional characterization of gradient flow dynamics for the matrix factorization problem, based on advanced tools from random matrix theory.